The in vitro micronucleus assay using imaging flow cytometry and deep learning.

Area of Science:

• Toxicology
• Genetics
• Biotechnology

Background:

• The in vitro micronucleus (MN) assay is a standard regulatory requirement for assessing chemical genotoxicity.
• Current scoring methods for the MN assay, including manual microscopy, automated microscopy, and flow cytometry, have limitations.
• Previous attempts using imaging flow cytometry (IFC) with feature-based analysis required extensive optimization.

Purpose of the Study:

• To develop a fully automated deep-learning method for scoring the MN assay using IFC data.
• To compare the performance of a deep-learning approach with existing methods, including manual scoring and feature-based analysis.
• To implement a robust and adaptable strategy for MN assay scoring across different cell types and chemicals.

Main Methods:

• Utilized imaging flow cytometry (IFC) with the ImageStream® platform for high-throughput image acquisition.
• Developed a deep-learning model utilizing convolutional neural networks (CNNs) via Amnis® AI software for image analysis.
• Scored both cytokinesis-blocked and unblocked versions of the MN assay using the developed deep-learning method.

Main Results:

• The deep-learning method demonstrated comparable performance to manual microscopy scoring.
• The Amnis® AI-based deep-learning approach outperformed the previous feature-based analysis in the IDEAS® software.
• Full automation of the MN assay scoring was achieved, reducing the need for complex optimization.

Conclusions:

• Deep learning with Amnis® AI offers a powerful and automated solution for scoring the in vitro MN assay.
• This novel approach enhances the efficiency and reliability of genetic toxicity testing.
• The method facilitates broader adoption and application of the MN assay in regulatory toxicology.